|Publication number||US4772445 A|
|Application number||US 06/812,202|
|Publication date||Sep 20, 1988|
|Filing date||Dec 23, 1985|
|Priority date||Dec 23, 1985|
|Also published as||EP0227479A2, EP0227479A3, EP0227479B1|
|Publication number||06812202, 812202, US 4772445 A, US 4772445A, US-A-4772445, US4772445 A, US4772445A|
|Inventors||Chaoukat N. Nasrallah, Kingsley F. Graham|
|Original Assignee||Electric Power Research Institute|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (18), Non-Patent Citations (4), Referenced by (18), Classifications (14), Legal Events (7)|
|External Links: USPTO, USPTO Assignment, Espacenet|
Xj =(1-W)Xj-1 +WXj,
N2 i,j =(1-Wn)N2 i,j-1 +Wn N2 i,j,
1. Field of the Invention
The present invention relates in general to signal validation systems, and more specifically to a signal validation system in the control system of a nuclear reactor; the signal validation system being capable of determining direct current (DC) drift and noise components in signal discrepancies.
2. Description of the Related Art
Reliability is an important consideration in many areas of technology, such as advanced avionic systems and the control systems of nuclear reactors. The use of redundant components, at least two, and more typically three or four components, is one technique that is used to increase reliability. The advantage of redundant components is that when a failure of one of the components is detected, the remaining component(s) which produce signals conveying the same information can still be used. Thus, system reliability becomes a function of the ability to identify and handle component failures.
In the case of signals output by sensors, several types of failures are possible, including: bias--due to D.C. drift or noise, hardover--to a maximum or minimum level; sticking--at a given level; or common mode--where more than one sensor fails in the same manner, more or less simultaneously. In the case of sensors which output positive and negative values corresponding to the maximum and minimum values of a hardover error, null failure, resulting in the output of a zero value, is also possible.
A powerful tool for identifying many types of failures while requiring relatively little variation for specific control systems, is provided by the parity-space algorithm. The parity-space algorithm was originally developed for advanced avionic systems as described in Potter and Suman, "Thresholdless Redundancy Management with Arrays of Skewed Instruments," Integrity in Electronic Flight Control Systems, NATO AGARDOGRAPH-224, 1977, pages 15-1 to 15-25 incorporated by reference herein. The application of the parity-space algorithm to control of nuclear reactors is described in a report available from the Electric Power Research Institute, Inc. as the Final Report dated Nov. 19, 1981 of NP-2110, Research Project 1541, by Deckert et al., entitled On-line Power Plant Signal Validation Technique Utilizing Parity-Space Representation and Analytic Redundancy incorporated by reference herein. As described therein, application of the parity-space algorithm to a specific system only requires determination of an error boundary for each sensor in the system.
The parity-space algorithm provides information concerning discrepancies between redundant measurements of a parameter. Rather than comparing the measurements against a reference value, the parity-space algorithm looks at the differences between the values. As a result, the number of dimensions q of a parity vector generated by the parity-space algorithm is equal to the number of measurements of a parameter minus the dimensions of the parameter. For example, according to the parity-space algorithm, l measurements of a scalar parameter produce a parity vector having q=l-1 dimensions.
One particular type of parity space has been termed "orthogonal parity space" by Potter and Suman in their article referenced above. As defined therein, orthogonal parity space has the following properties. An orthogonal parity vector p is defined according to equation (1), where V is an upper triangular matrix that transforms l measurements in an l by 1 column vector m into the parity vector p.
The measurement vector m is defined according to equation (2), where x is the actual value of the parameter being measured and has n dimensions where n equals 1 if the parameter being measured is a scaler, ε is the error in each of the measurements, and H is an l by n matrix.
Thus, when a scalar is measured by three sensors, H is defined by equation (3). ##EQU1## In orthogonal parity space, matrix V is defined to have the properties of equation (5) where matrix K is defined as in equation (4).
K=(HT H)-1 HT ( 4)
VT V=I-HK=W (5)
From equation (5) it follows that equations (6) and (7) are true.
VVT =KH=I (6)
VH=KVT =0 (7)
Given the above characteristics of matrix V, Potter and Suman have found that the elements of the matrix V are defined by equations (8)-(12). ##EQU2##
The simplest application of the parity-space algorithm in orthogonal parity space occurs when a scalar parameter (n=1) is measured by three sensors (l=3). In this case, the matrix H is defined by equation (3) and, as defined in equation (4), K=1/3[1,1,1]. Using the definition in equation (5), the elements of matrix W have the values in equation (13). ##EQU3## The formulas in equation (8)-(12) give the following results for matrix V. ##EQU4## The columns of matrix V are vectors of length √2/3 which lie along the measurement axes in parity space. Plotting these vectors results in a positively valued measurement axis every 120°, as illustrated in FIG. 1.
As a result of the special properties of orthogonal parity space, several meaningful values can be found using a minimum amount of calculation when the parameter being measured is a scalar. A residual ηj can be found for each of the measurements mj by subtracting the mean m from each measurement mj, i.e., ηj =mj -m, where m is calculated according to equation (15). ##EQU5## The residuals can then be reordered from smallest to largest, as indicated in equation (16).
η1 ≦η2 ≦ . . . ≦ηl ( 16)
After the reordering, there is a one-to-one correspondence between the measurements mj and the residuals ηj, but the residual η2, for example, is not necessarily the residual for the measurement m2. After such reordering, a projection pj of the parity vector along the measurement axis of each sensor j can be can be calculated according to equation (17).
Pj =√l/(l-1)·ηj ( 17)
Given measurements from three sensors, for example, the parity vector 10 will be two-dimensional and thus easily depicted on a display screen 20, as illustrated in FIG. 1. In the case of such a two-dimensional parity vector 10, it is relatively easy to convert the projection pj into two-dimensional component parity vectors (the coordinates of the parity vector in the plane), or into sensor components of the parity vector in the direction of the measurement axes, using appropriate geometric and trigonometric relationships. An example of conversion will be given in the Description of the Preferred Embodiment.
Failure of any one of the sensors can be detected by analyzing the parity vector. When the sensors are assumed to have a uniform error boundary b, at least one of the sensors is in error if the inequality (18) is satisfied, where δa is defined by equations (19a) and (19b). ##EQU6##
In many cases, it is possible to identify which of the sensors has failed by calculating orthogonal components ##EQU7## in accordance with equation (20) for each of the sensors. ##EQU8## Since the residuals ηj were ordered according to equation (16) above, the orthogonal component ##EQU9## will have the largest value of any of the orthogonal components. Therefore, if the orthogonal component ##EQU10## is small, i.e., the inequality in equation (21), where δl-1 is defined by equations (19a) and (19b), is true for j=1, then no inconsistency has been detected in the measurements supplied by the sensors and the value of the parameter measured by the sensors can be estimated as the mean m. ##EQU11## If the inequality in equation (21) is not satisfied for j=1, then the inequality is tested repeatedly for each value of j=2 through j=l. As j gets larger, the value of p2 j will get smaller. If the inequality in equation (7) is satisfied for a value of j between 2 and l-1, inclusive, then the sensor which produced the inconsistency cannot be positively identified, and the best estimate for the value of the parameter measured by the sensors is m, where m is defined by equation (22). ##EQU12## If the first value of j which satisfies the inequality in equation (21) is l, then the sensor which generated the residual ηl can be identified as having produced the inconsistent measurement and equation (22), with j=l, can be used for the best estimate m of the value of the parameter measured by the sensors.
If none of the measurements can be identified as having been inconsistent, then the above procedure is repeated, throwing out the measurement which generated the largest residual ηl. This is equivalent to decrementing the value of l by 1 and repeating equations (15) through (22). However, if the value of l is decremented below three, i.e., only two measurements are left, without isolating an inconsistent measurement, it is impossible to isolate the inconsistent measurement using the conventional parity-space algorithm.
Assuming this does not occur, when the equation in (21) is satisfied and the value of the decremented l is at least three, then the estimate m for the value of the parameter measured by the sensors can be calculated according to equation (22). However, if the value of j which satisfies the inequality in equation (21) equals the value of the decremented l, it is necessary to compare the measurements of the excluded sensors with the measurement of the sensor having the orthogonal component ##EQU13## which satisfied equation (21) to determine whether the difference therebetween is less than the error-bound b. In other words, if the inequality in equation (23) is true for any value of k greater than the value of j and less than or equal to the original value of l, where the value of j equals the value of the decremented l, then a common mode inconsistency has occurred which the parity-space algorithm is unable to isolate.
|mk -mj |≦2b (23)
When a sensor is identified as having generated consecutive inconsistent measurements, for example, three times, then that sensor is identified as having failed.
As described above, although the basic parity-space algorithm is extremely powerful, it has limitations with respect to the types of failures which are detected. For example, a sensor may have a significant amount of DC drift, e.g., due to poor calibration, without having actually failed. Since the parameter-space algorithm does not distinguish between noise and DC drift contributions to inconsistent measurements, the information which could be provided by analysis of these components is not utilized.
An object of the present invention is to provide a system for determining direct current drift and noise levels in redundant sensors using parity-space signal validation.
A further object of the present invention is to identify one or more failed sensors in situations in which parity-space signal validation is unable to identify a failed sensor.
Another object of the present invention is to provide a direct current drift signal for each sensor in a system utilizing redundant sensors to permit remote calibration of the sensors.
The above mentioned objects are attained by utilizing a method of measuring direct current drift and noise in samples of sensor signals received from corresponding sensors, including the step of converting the samples of the sensor signals into parity vector signals, each of the parity vector signals representing inconsistancy among a corresponding sample of the sensor signals. The method also includes the steps of converting the parity vector signals into a direct current drift signal by producing a first running average of the parity vector signals and producing an instantaneous noise signal for a corresponding parity vector signal by subtracting the direct current drift signal from the corresponding parity vector signal.
These objects, together with other objects and advantages which will be subsequently apparent, reside in the details of construction and operation as more fully hereinafter described and claimed, reference being had to the accompanying drawings forming a part hereof, wherein like reference numerals refer to like parts throughout.
FIG. 1 is a planar representation of parity space illustrating calculation of DC drift;
FIG. 2 is a block diagram of an apparatus to which the present invention can be applied;
FIG. 3 is a planar representation of parity space illustrating calculation of sensor DC drift vectors;
FIG. 4 is a planar representation of parity space indicating the meaning of vectors in parity space;
FIG. 5, is a planar representation of parity space illustrating calculation of a composite noise vector; and
FIG. 6, is a planar representation of a region in parity space illustrating calculation of sensor noise vectors.
Conventionally, control systems utilize sensors 30 (FIG. 2) to measure critical variables in a system controlled by a control system 35. The system being controlled may be in an aircraft or spacecraft, a chemical manufacturing plant, etc. The present invention can be applied to any such system, but will be described below for sensors 30 in a pressurized light water nuclear reactor. The sensors 30 provide sensor signals via lines 40 to an analog/digital converter 45, such as the A/D converter on an Intel 88/40 single board computer which can provide some preprocessing. Even using the best shielding available, the sensor signals received by the analog/digital converter 45 will contain some amount of noise since the sensors typically see process noise in the parameter they are monitoring. An example is steam generation level noise caused by the boiling in the steam generator. In addition, the sensor signals may contain direct current (DC) drift due to: poor initial calibration of the sensors 30, drifting of the sensors 30 during operation or failure of the sensors 30.
The present invention identifies DC drift and noise signals present in the sensor signals by analyzing digitized sensor signals output by the analog/digital converter 45 in a processor 50, such as an Intel 86/05. The results of the analysis can be displayed on an auxiliary display 55 or used to control the nuclear reactor control system 35 after conversion to analog signals in a digital/analog converter 60.
The first step in measuring direct current drift and noise according to the present invention is to generate component parity vector signals as described above with reference to equations (1) through (17). As illustrated in FIG. 1, the parity vectors define coordinates 100 that may be limited by an envelope 110 which defines the limit of the noise seen by a long sequence of measurements. Averaging the coordinates 100 over time produces a centroid 120 which corresponds to composite direct current drift, since the effects of noise will be averaged out. The centroid 120 can be calculated using standard matrix arithmetic.
Given the coordinates of the centroid 120, it is possible to generate sensor direct current drift signals in the direction of the q or (l-1) measurement axes 130, 140 and 150 corresponding to sensors #1, #2 and #3, respectively. Since, as described above, the parity vector has l-1 dimensions when there are l sensors, the maximum number of constituent sensors corresponding to the sensor direct current drift signals is l-1. An example will be given below for l equals three, since the resulting parity vector will have two dimensions and can be easily illustrated. However, the number of sensors which can be handled by the present invention is not limited to three. If only two sensors are available, it may be possible to calculate one or more estimated sensor signals by assuming certain relationships between other measured parameters and the parameter being measured by the two remaining sensors (see the EPRI report on Research Project 1541 referenced above).
One method for identifying the constituent sensors is to consider equations (16) and (17). Since the direct current drift signal is simply an average parity vector signal, if an average residual ηj is kept for each of the sensors, the sensor producing the smallest average residual η1 will produce the smallest projection p1 according to equation (17). In other words, the sensor producing the smallest average residual η1 contributes the least to the DC drift signal and is, therfore, not a constituent sensor.
The first method of identifiying constituent sensors may be better understood in view of a second method for identifying the contituent sensors of the DC drift signal. The second method is most easily explained by considering the graphical representation of two-dimensional parity. In the case of a two-dimensional composite DC drift signal or vector 155, defined by the centroid 120 and the origin 156, there are at most two constituent sensors which may have corresponding component DC drift signals. The constituent sensors can be identified by the angle between the vector 155 and a given measurement axis.
Given the definition of matrix V, in particular equation (9) which defines an upper triangular matrix, the first column has only a single non-zero element V11 and thus can be easily assigned to the X-axis. Thus, in FIG. 3, the sensor signal provided by sensor number 1 is assigned to the X-axis 130. The matrix elements in the sensor signals from sensors 2 and 3 are thus displayed along axes 140 and 150 forming +120° and -120° angles with the positive X-axis, respectively. The angle φ formed between vector 155 and the X-axis 130 identifies the constituent sensors. If the angle φ is greater than zero and less than 60°, then the parity vector signal of DC drift signal will include a sensor signal from sensor number 1 with a positive deviation from the average and a sensor signal from sensor number 3 with a negative deviation, as illustrated in FIG. 3. Similarly, positive values of angle φ between 60° and 120°, and 120° and 180° result in constituent sensors 2 and 3, and 1 and 2, respectively, while negative values of φ between 0° and -60°, -60° and -120°, and -120° and -180° have corresponding constituent sensors 1 and 2, 2 and 3, and 1 and 3, respectively, as illustrated in FIG. 3.
After the constituent sensors have been identified, sensor DC drift signals can be found for each of the constituent sensors using geometric and trigonometric relationships. The case of three measurements of a scalar parameter will be used. Two examples of the computational methods will be given, but it is should be understood that others are possible. In the first method, it is noted that the measurement axis 150 of sensor #3 is defined by y=√3 x (see the third column of matrix in equation (14)), while y=-√3 x defines the measurement axis 140 of sensor #2. In four out of the six regions defined in FIG. 3, sensor #1 is a constituent sensor and the sensor #1 DC drift can easily be found by finding the X-intercept of a line 160 (FIG. 1) parallel to the measurement axis of the other constituent sensor (#2 or #3). The other sensor (#2 or #3) DC drift has a Y-coordinate equal to the Y-coordinate of the composite DC drift and its X-coordinate can be found by solving the appropriate equation above defining the measurement axis for that sensor. Thus, the component DC drift signals 210 and 220 in FIG. 1 can be easily found.
In the case of a composite DC drift signal in a region in which the constituent sensors are 2 and 3, calculation of the component DC drift signals is slightly more difficult, but the same principles are involved. See for example, the composite DC drift vector 155' in FIG. 3 in a region having constituent sensors #2 and #3.
A second method of calculating sensor DC drift signals utilizes the residual signals corresponding to each of the sensors. As noted above, the projection along each of the measurement axes is defined by equation (17). However, √l/(l-1) is simply a scale factor which puts calculations using the residuals into the same scale as those using matrix V, defined in equation (14). It is possible to perform the calculations directly on residuals ηj and multiply by the scale factor √l/(l-1) at a later time or perform scaling using some other factor.
With reference to FIG. 3, the coordinates of the vector 155' representing composite DC drift and the sensor #2 and sensor #3 DC drift signals represented by vectors 220' and 230' can be calculated from the projections pj along the measurement axes, as defined by equation (17). By definition, a line 170 between the tip of the DC drift vector 10' and the tip p3 of the projection along the sensor #3 measurement axis 150 forms an angle of 30° with the measurement axis 130, since the smallest angle between all of the measurement axes if 60°. Therefore, X=-p3 /sin 30°. The X-coordinate XDC equals the projection p1 along the sensor #1 measurement axis 130 and the Y-coordinate YDC of the composite DC vector 10' is defined by equation (24).
YDC =(p1 -X) tan 30° (24)
The sensor #2 and #3 DC drift signals represented by vectors 220' and 230' can be calculated from the composite DC drift vector 155' utilizing the parallelogram having sides formed by vectors 220' and 230' and having a major diagonal formed by vector 155', together with the triangle defined by the tips of vectors 155', 220' and p3. The distance D between p3 and the tip of vector 10' can be found by vector subtraction or as D=p3 cos θ, where θ=120°-arctan YDC /XDC. Thus, the sensor #2 DC drift signal DC2 =-D/sin 60° and the sensor #3 DC drift signal DC3 =p3 -DC2 cos 60°. When one of the consituent sensors of the composite DC drift signal is sensor #1, the sensor DC drift signals are easily calculated. For example, the sensor #1 and #3 DC drift signals DC3 and DC1 represented by vectors 210 and 220 (FIG. 1), respectively, are found as DC3 =-YDC /sin 60° and DC1 =XDC +DC3 cos 60°.
Increased sensitivity to sudden changes in DC drift can be provided according to the present invention by weighting the average which produces the centroid 120. Such weighting is provided by equation 25, where Xi,j is the weighted running average for sensor direction i of the parity vector in the jth sample of the sensor signals, (1-W) is a weighting coefficient in which W is a weighting function and Xi,j is the sensor #i parity vector signal for in the jth sample. The weighting function W may, for example, be 0.01 or 0.02, representing an average over the last 100 to 50 samples, respectively.
Xi,j =(1-W)Xi,j-1 +W·Xi,j (25)
Equation (25) represents first order lag, but other equations may be used to provide a desired response, representing, e.g., second order lag, for recent samples of the sensor signals, as is known the art. Other variations are also possible. For example, the parity vector can be averaged in two dimensions and periodically (e.g., every 50 or 100 samples) the sensor components can be calculated, as described above.
As noted above, noise on lines 40 (FIG. 2) cause fluctuations in the parity vector corresponding to the sample of the sensor signals. Thus, in a recent sample of the sensor signals, a parity vector 310 (FIG. 5) may be produced having coordinates which are different from the coordinates of the composite DC drift vector 155. Using standard vector arithmetic, a composite instaneous noise signal can be calculated and represented by vector 320.
The composite instaneous noise signal has sensor instaneous noise signals corresponding to constituent sensors. A magnified drawing of the region on display screen 20 bounded by positive values of sensor #1 and negative values of sensor #3 is illustrated in FIG. 6. Component instaneous noise signals corresponding to constituent sensors and represented by vectors 340 and 350 can be found by transposing the axes 130, 140 and 150 to the centroid 120 to produce a coordinate system for noise signals defined by axes 130', 140' and 150', corresponding to sensors #1, #2 and #3, respectively. After the constituent sensors of the noise signal 230 are identified, the sensor instantaneous noise signals represented by vector 340 and 350 can be found using the same methods used to find the sensor DC drift signals.
The average amount of noise in each sensor signal can be found using a weighted RMS average of the RMS instantaneous noise signals. Thus, a sensor noise signal N2 k,j can be calculated using standard matrix arithmetic and equation (26), where N2 k,j is the sensor noise signal for sensor number k in the jth sample of the sensor signals, Wn is a weighting function for noise signals which may or may not be the same as the weighting function W for DC drift signals, Xj is the parity vector signal for the jth sample of the sensor signals, Xj is the direct current drift vector, and k is a unit vector along the axis for sensor number k.
N2 k,j =(1-Wn)N2 k,j-1 +Wn [(Xj -Xj)·k]2 (26)
The lines parallelling each of the axes 130, 140 and 150 in FIGS. 1 and 2 indicate the error bounds of each of the sensors. In the above example, it has been assumed that the error bound for each of these sensor signals is the same and has a value of b, as used in equations (4), (5a), (5b) and (9). However, it is possible for the error bounds of the sensors to differ and this can be taken into account in the parity-space algorithm.
The many features and advantages of the present invention are apparent from the detailed specification, and thus it is intended by the appended claims to cover all such features and advantages of the system which fall within the true spirit and scope of the invention. Further, since numerous modifications and changes will readily occur to those skilled in the art, it is not desired to limit the invention to the exact construction and operation illustrated and described, accordingly, all suitable modifications and equivalents may be resorted to falling within the scope and spirit of the invention.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US3568150 *||Apr 16, 1969||Mar 2, 1971||United Aircraft Corp||Noise discriminating fault apparatus|
|US3667057 *||May 22, 1970||May 30, 1972||Bendix Corp||Method and means for providing an output corresponding to the average of acceptable input signals|
|US3764467 *||Dec 22, 1971||Oct 9, 1973||Atomic Energy Commission||Fast breeder reactor protection system|
|US3770948 *||May 26, 1972||Nov 6, 1973||Gte Automatic Electric Lab Inc||Data handling system maintenance arrangement|
|US3881670 *||Dec 7, 1973||May 6, 1975||Bendix Corp||Control system with dual unmonitored condition sensors and fail operative characteristics|
|US4105496 *||Dec 5, 1975||Aug 8, 1978||Commissariat A L'energie Atomique||Method and device for electronic control with positive safety|
|US4223681 *||Apr 10, 1978||Sep 23, 1980||Hewlett-Packard Company||Validation of blood pressure|
|US4276648 *||Sep 4, 1979||Jun 30, 1981||The Boeing Company||Midvalue signal selection and fault detection apparatus and method|
|US4327437 *||Jul 30, 1980||Apr 27, 1982||Nasa||Reconfiguring redundancy management|
|US4346363 *||Dec 22, 1978||Aug 24, 1982||Jack Wilhelmson||Electronic circuit for sensing deviations in physical forces thereon|
|US4433414 *||Sep 30, 1981||Feb 21, 1984||Fairchild Camera And Instrument Corporation||Digital tester local memory data storage system|
|US4472806 *||May 3, 1982||Sep 18, 1984||The Boeing Company||Signal selection and fault detection apparatus|
|US4497059 *||Apr 28, 1982||Jan 29, 1985||The Charles Stark Draper Laboratory, Inc.||Multi-channel redundant processing systems|
|US4594714 *||May 2, 1983||Jun 10, 1986||United Technologies Corporation||Dual-actuator monitor|
|US4625314 *||Jun 5, 1984||Nov 25, 1986||Siemens Aktiengesellschaft||Method for the recognition of the outage of one or more transmission channels in a redundantly designed optical transmission system|
|US4654582 *||Jul 8, 1985||Mar 31, 1987||The United States Of America As Represented By The Secretary Of The Air Force||Transient test of suspension electronics for gyroscope|
|US4667284 *||May 5, 1984||May 19, 1987||Hitachi, Ltd.||Multiplexing control unit|
|DE2636352A1 *||Aug 12, 1976||Feb 16, 1978||Kraftwerk Union Ag||Reactor protection system with data processing of readings - has second computer parallel to first to improve reliability|
|1||Deckert et al., "On Line Power Plant Signal Validation Technique Utilizing Parity-Space Representation and Analytic Redundancy," Electric Power Research Institute, Final Report of Research Project 1541, Nov. 19, 1981.|
|2||*||Deckert et al., On Line Power Plant Signal Validation Technique Utilizing Parity Space Representation and Analytic Redundancy, Electric Power Research Institute, Final Report of Research Project 1541, Nov. 19, 1981.|
|3||Potter and Suman, "Thresholdless Redundancy Management with Arrays of Skewed Instruments," Integrity in Electronic Flight Control Systems, NATO AGARDOGRAPH-224, 1977, pp. 15-1 to 15-25.|
|4||*||Potter and Suman, Thresholdless Redundancy Management with Arrays of Skewed Instruments, Integrity in Electronic Flight Control Systems, NATO AGARDOGRAPH 224, 1977, pp. 15 1 to 15 25.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US5267277 *||Nov 2, 1989||Nov 30, 1993||Combustion Engineering, Inc.||Indicator system for advanced nuclear plant control complex|
|US5410492 *||May 27, 1993||Apr 25, 1995||Arch Development Corporation||Processing data base information having nonwhite noise|
|US5459675 *||May 27, 1993||Oct 17, 1995||Arch Development Corporation||System for monitoring an industrial process and determining sensor status|
|US5586066 *||Jun 8, 1994||Dec 17, 1996||Arch Development Corporation||Surveillance of industrial processes with correlated parameters|
|US5629872 *||Apr 8, 1996||May 13, 1997||Arch Development Corporation||System for monitoring an industrial process and determining sensor status|
|US5661735 *||Dec 26, 1995||Aug 26, 1997||Litef Gmbh||FDIC method for minimizing measuring failures in a measuring system comprising redundant sensors|
|US5761090 *||Oct 10, 1995||Jun 2, 1998||The University Of Chicago||Expert system for testing industrial processes and determining sensor status|
|US6073262 *||May 30, 1997||Jun 6, 2000||United Technologies Corporation||Method and apparatus for estimating an actual magnitude of a physical parameter on the basis of three or more redundant signals|
|US6356857||Oct 27, 1998||Mar 12, 2002||Aspen Technology, Inc.||Sensor validation apparatus and method|
|US6594620||Dec 29, 1999||Jul 15, 2003||Aspen Technology, Inc.||Sensor validation apparatus and method|
|US7451003||Mar 4, 2004||Nov 11, 2008||Falconeer Technologies Llc||Method and system of monitoring, sensor validation and predictive fault analysis|
|US8837664 *||Oct 24, 2013||Sep 16, 2014||Kabushiki Kaisha Toshiba||Power monitoring system|
|US20050113771 *||Nov 26, 2003||May 26, 2005||Kimberly-Clark Worldwide, Inc.||Odor control in personal care products|
|US20100254504 *||Sep 29, 2009||Oct 7, 2010||Kabushiki Kaisha Toshiba||Power Monitoring System|
|DE4100501A1 *||Jan 10, 1991||Jul 16, 1992||Bodenseewerk Geraetetech||Detecting and identifying faults at sensors for state values - using association matrix giving optimal relationship between characteristic vectors and associated classification vectors|
|WO1994028557A1 *||May 27, 1994||Dec 8, 1994||Arch Dev Corp||System for monitoring an industrial process and determining sensor status|
|WO2000010059A1 *||Oct 27, 1998||Feb 24, 2000||Aspen Technology Inc||Sensor validation apparatus and method|
|WO2006005858A1 *||May 12, 2005||Jan 19, 2006||Renault Sa||Method of monitoring one or more physical parameters and fuel cell using same|
|U.S. Classification||376/245, 714/817, 714/800, 367/126, 702/195, 376/217, 318/564, 367/124, 318/563|
|International Classification||G01D21/00, G21C17/00, G01D3/02|
|Dec 23, 1985||AS||Assignment|
Owner name: WESTINGHOUSE ELECTRIC CORPORATION, WESTINGHOUSE BU
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST.;ASSIGNORS:NASRALLAH, CHAOUKAT N.;GRAHAM, KINGSLEY F.;REEL/FRAME:004497/0449
Effective date: 19851210
|Mar 17, 1987||AS||Assignment|
Owner name: EPRI
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST. SUBJECT TO LICENSE RECITED.;ASSIGNOR:WESTINGHOUSE ELECTRIC CORPORATION, A PA. CORP.;REEL/FRAME:004682/0248
Effective date: 19870108
|Nov 20, 1991||FPAY||Fee payment|
Year of fee payment: 4
|Apr 30, 1996||REMI||Maintenance fee reminder mailed|
|Jun 14, 1996||FPAY||Fee payment|
Year of fee payment: 8
|Jun 14, 1996||SULP||Surcharge for late payment|
|Feb 25, 2000||FPAY||Fee payment|
Year of fee payment: 12